Review of neural network applications in medical imaging and signal processing

Abstract

The current applications of neural networks to in vivo medical imaging and signal processing are reviewed. As is evident from the literature neural networks have already been used for a wide variety of tasks within medicine. As this trend is expected to continue this review contains a description of recent studies to provide an appreciation of the problems associated with implementing neural networks for medical imaging and signal processing.

Artificial Neural Networks (ANN) Applied for Gait Classification and Physiotherapy Monitoring in Post Stroke Patients

“…with one hidden layer); here data concerning gait rhythm, speed and five kinetic values were fed into the neural network, based on which four gait types were correctly distinguished at a rate of 80%. The review of Miller et al (1992) described other neural networks applied in different areas of investigation, with similar results. In turn, Aminian et al (1995) used accelerometer to recognize the speed and incline of walking.…”

Review of neural network applications in sleep research

“…This approach is characteristic of the great majority of models (in comparison with other connectionist models) used in the field of biological data processing (Kemsley et al, 1991;Leuthausser, 1991;Miller et al, 1992;Micheli-Tzanakou, 1995). Another important similarity is the difficulty encountered by Schaltenbrand et al (1993) and Pfurtscheller et al (1992) in discriminating sleep stages 3 from sleep stage 4 in human.…”

“…Neural networks are generally regarded as black boxes. This characteristic is sometimes considered a drawback and put forward to restrain their utilization as a clinical diagnosis support, but results obtained with the connectionist systems (Kemsley et al, 1991;Miller et al, 1992;Itchhaporia et al, 1996) justify their use.…”

“…Over several decades, thanks to the development of microcomputer technologies, automated systems to build hypnograms have emerged, either based on classical algorithms (Gaillard and Tissot, 1973;Lim and Winters, 1980;Itowi et al, 1990;Witting et al, 1996), or artificial intelligence methods (Kumar, 1977;Ray et al, 1986). Neural networks (or connectionist models) are a new form of artificial intelligence techniques recently introduced to biological data processing (Gevins and Morgan, 1988;Kemsley et al, 1991;Leuthausser, 1991;Miller et al, 1992;Micheli-Tzanakou, 1995).…”

Validation Procedures in Radiological Diagnostic Models. Neural Network and Logistic Regression

“…Lesions were divided into benign (0) and malignant (1). Out of the 19 explanatory variables, three were continuous (1,4,8), one was quantitative discrete (5), two were ordinal (11,12) and the rest were qualitative. For these latter, we used a 1-out-of C code where a variable with C categories is converted in C Boolean inputs, each of which is high for a certain category; eventually 43 explanatory variables were used.…”